CN116383661A - Centrifugal pump fault diagnosis model training method, fault diagnosis method and device - Google Patents
Centrifugal pump fault diagnosis model training method, fault diagnosis method and device Download PDFInfo
- Publication number
- CN116383661A CN116383661A CN202310417097.5A CN202310417097A CN116383661A CN 116383661 A CN116383661 A CN 116383661A CN 202310417097 A CN202310417097 A CN 202310417097A CN 116383661 A CN116383661 A CN 116383661A
- Authority
- CN
- China
- Prior art keywords
- centrifugal pump
- time domain
- fault diagnosis
- acoustic radiation
- feature matrix
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 117
- 238000000034 method Methods 0.000 title claims abstract description 94
- 238000012549 training Methods 0.000 title claims abstract description 67
- 239000011159 matrix material Substances 0.000 claims abstract description 78
- 230000005855 radiation Effects 0.000 claims abstract description 69
- 238000013528 artificial neural network Methods 0.000 claims abstract description 45
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 33
- 238000012545 processing Methods 0.000 claims abstract description 18
- 230000002708 enhancing effect Effects 0.000 claims abstract description 14
- 230000009467 reduction Effects 0.000 claims description 16
- 238000009499 grossing Methods 0.000 claims description 12
- 238000003062 neural network model Methods 0.000 claims description 8
- 238000012847 principal component analysis method Methods 0.000 claims description 7
- 210000002569 neuron Anatomy 0.000 claims description 6
- 230000008878 coupling Effects 0.000 claims description 2
- 238000010168 coupling process Methods 0.000 claims description 2
- 238000005859 coupling reaction Methods 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 238000005096 rolling process Methods 0.000 claims description 2
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 239000013598 vector Substances 0.000 description 14
- 238000012360 testing method Methods 0.000 description 11
- 230000000694 effects Effects 0.000 description 9
- 241000283973 Oryctolagus cuniculus Species 0.000 description 8
- 238000010586 diagram Methods 0.000 description 8
- 230000006870 function Effects 0.000 description 7
- 238000000605 extraction Methods 0.000 description 6
- 238000009434 installation Methods 0.000 description 6
- 230000009286 beneficial effect Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 4
- 238000005457 optimization Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000001514 detection method Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000000750 progressive effect Effects 0.000 description 2
- 238000013139 quantization Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 206010033799 Paralysis Diseases 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 230000009189 diving Effects 0.000 description 1
- 239000012636 effector Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 239000011664 nicotinic acid Substances 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Structures Of Non-Positive Displacement Pumps (AREA)
Abstract
The invention discloses a centrifugal pump fault diagnosis model training method, a centrifugal pump fault diagnosis method and a device, and relates to the field of centrifugal pump fault diagnosis, wherein the centrifugal pump fault diagnosis model training method comprises the following steps: acquiring centrifugal pump acoustic radiation signals under the same working condition and different health states; enhancing the centrifugal pump acoustic radiation signal, and processing the enhanced centrifugal pump acoustic radiation signal to obtain a first effective time domain feature matrix; and taking the first effective time domain feature matrix as a sample data set, and optimizing a probability neural network through a Harris eagle algorithm to form a centrifugal pump fault diagnosis model. The invention realizes that the fault type of the centrifugal pump can still be identified efficiently and accurately under the interference of strong background noise, and has important significance for early warning of the fault of the centrifugal pump.
Description
Technical Field
The invention relates to the technical field of centrifugal pump fault diagnosis, in particular to a centrifugal pump fault diagnosis model training method, a centrifugal pump fault diagnosis method and a centrifugal pump fault diagnosis device.
Background
The centrifugal pump is widely applied in the industrial field, is one of key equipment in the fields of petrochemical industry, military national defense and the like, and is light in failure, thereby causing paralysis of a production chain, and heavy in safety accident, thereby causing casualties. Therefore, the method has extremely important significance for rapid and accurate fault diagnosis of the centrifugal pump.
Currently, for fault diagnosis during the operation of a centrifugal pump, the following two methods are generally adopted:
1. the method based on vibration signals generally adopts a contact sensor to collect vibration signals generated when the centrifugal pump operates as an analysis basis, but in actual working conditions, on one hand, the pump group is compact in installation and arrangement, the space is narrow, on the other hand, the operation working conditions are complex, a high-temperature environment exists, and non-contact vibration measurement is difficult to be carried out on the centrifugal pump by using the vibration sensor.
2. Based on the model method, various fault signals are calculated through simulation by establishing a mathematical model of the rotary mechanical system, the information is compared with vibration information obtained through actual monitoring, and fault diagnosis is carried out through analysis of residual errors. However, since it is difficult to build an accurate mathematical model in the rotary mechanical device, complicated parameter adjustment and comparison analysis are required in the fault detection of the centrifugal pump, which increases the workload and cannot ensure the accuracy of the fault detection.
Therefore, how to quickly, efficiently and accurately detect the faults of the centrifugal pump is a problem to be solved at present.
Disclosure of Invention
In view of the above, the invention provides a centrifugal pump fault diagnosis model training method, a centrifugal pump fault diagnosis method and a centrifugal pump fault diagnosis device, so as to solve the problems of complex operation and low accuracy in the prior art.
In a first aspect, an embodiment of the present application provides a method for training a fault diagnosis model of a centrifugal pump, where the method includes:
acquiring sound radiation signals of the centrifugal pump under different states;
enhancing the centrifugal pump acoustic radiation signal, and processing the enhanced centrifugal pump acoustic radiation signal to obtain a first effective time domain feature matrix;
and taking the first effective time domain feature matrix as a sample data set, and optimizing a probability neural network through a Harris eagle algorithm to form a centrifugal pump fault diagnosis model.
In an embodiment of the present application, the acquiring acoustic radiation signals of the centrifugal pump under different states includes:
disposing a microphone adjacent the centrifugal pump;
and setting different fault types of the centrifugal pump, and collecting centrifugal pump acoustic radiation signals in different fault states and centrifugal pump acoustic radiation signals in a normal state through the microphone.
In an embodiment of the present application, the different fault types include: the motor shaft is not centered with the shaft axis of the centrifugal pump, the anchor bolts of the motor and the centrifugal pump are loosened, the centrifugal pump is cavitation corroded, the rolling bearing of the centrifugal pump is failed, the mechanical seal of the centrifugal pump is leaked, and the centrifugal pump is in coupling failure with the motor.
In an embodiment of the present application, the enhancing the centrifugal pump acoustic radiation signal includes:
and carrying out signal enhancement on the centrifugal pump acoustic radiation signal by using a dislocation superposition method.
In an embodiment of the present application, the processing the enhanced acoustic radiation signal of the centrifugal pump to obtain a first effective time domain feature matrix includes:
calculating the time domain characteristics of the enhanced centrifugal pump acoustic radiation signals;
constructing the time domain feature as a time domain feature matrix;
performing dimension reduction on the time domain feature matrix by using a principal component analysis method;
and carrying out normalization processing on the time domain feature matrix after the dimension reduction to obtain a first effective feature matrix.
In an embodiment of the present application, the time domain features include: average, peak, variance, root mean square, crest factor, margin factor, pulse coefficient, shape coefficient, skewness, and kurtosis.
In an embodiment of the present application, the optimizing the probabilistic neural network by the harris eagle algorithm using the first effective time domain feature matrix as a sample data set specifically includes:
establishing a probabilistic neural network, inputting a training set in the sample data set, training a neural network model, and initializing the neuron numbers of an input layer, a mode layer and a summation layer;
initializing parameters of a Harris eagle algorithm, and setting initial population quantity, maximum iteration times, upper and lower boundaries of a global search range and errors;
calculating the fitness value of each Harisc eagle individual in the population according to the upper bound and the lower bound of the global search range, and sequencing to determine that the Harisc eagle at the optimal position is the first generation Harisc eagle;
performing updating strategy classification according to the escape energy factors and the value range of the random numbers, and performing iterative updating on the Harris eagle position according to different strategies;
calculating the updated position of the harris eagle as the fitness after the iteration, comparing the updated position with the harris eagle at the optimal position of the previous generation, and reserving the harris eagle at the optimal position;
when the fitness value reaches a set value or the iteration number of the algorithm reaches a preset value, training is stopped, otherwise, the Harris eagle position is continuously updated;
and obtaining the final Harris eagle position, and substituting the Harris eagle fitness value into a model as an optimal smoothing factor of the probabilistic neural network model.
In a second aspect, the present application also discloses a method for diagnosing a fault of a centrifugal pump, the method comprising:
collecting a centrifugal pump acoustic radiation signal;
enhancing the centrifugal pump acoustic radiation signal, extracting the time domain feature of the enhanced centrifugal pump acoustic radiation signal, constructing a time domain feature matrix, and performing dimension reduction clustering on the time domain feature matrix to obtain a second effective time domain feature matrix;
and inputting the second effective time domain feature matrix into a centrifugal pump fault diagnosis model for fault diagnosis, wherein the centrifugal pump fault diagnosis model is obtained through training by the centrifugal pump fault diagnosis model training method in the first aspect of the application.
In a third aspect, the present application further discloses a centrifugal pump fault diagnosis model training device, including:
sample acquisition module: the method is used for acquiring acoustic radiation signals of the centrifugal pump under different states;
sample enhancement module: the method comprises the steps of enhancing the centrifugal pump acoustic radiation signal, and processing the enhanced centrifugal pump acoustic radiation signal to obtain a first effective time domain feature matrix;
model training module: and the first effective time domain feature matrix is used as a sample data set, and a probabilistic neural network is optimized through a Harris eagle algorithm so as to form a centrifugal pump fault diagnosis model.
In a fourth aspect, the present application also discloses a centrifugal pump failure diagnosis device, which is characterized by comprising:
the signal acquisition module: the device is used for collecting the acoustic radiation signals of the centrifugal pump;
a signal enhancement module: enhancing the centrifugal pump acoustic radiation signal, extracting the time domain feature of the enhanced centrifugal pump acoustic radiation signal, constructing a time domain feature matrix, and performing dimension reduction clustering on the time domain feature matrix to obtain a second effective time domain feature matrix;
and a fault diagnosis module: the second effective time domain feature matrix is used for inputting the second effective time domain feature matrix into a centrifugal pump fault diagnosis model for fault diagnosis, wherein the centrifugal pump fault diagnosis model is obtained through training by the centrifugal pump fault diagnosis model training method according to the first aspect of the application.
The invention has the beneficial effects that:
according to the centrifugal pump fault diagnosis model training method, the centrifugal pump fault diagnosis method and the centrifugal pump fault diagnosis device, an improved centrifugal pump fault diagnosis model is provided, the optimization of a Harris eagle algorithm is achieved, the self-adaptive selection of an optimal smoothing factor is achieved, the problem that the classification result is uncertain due to the fact that the smoothing factor of a traditional probability neural network is manually adjusted is solved, the convergence speed and performance of the optimized network are good, and reliable fault diagnosis of the centrifugal pump is achieved; in addition, the invention constructs a non-contact fault diagnosis method and system based on acoustic radiation signals, and only needs to arrange the microphone on the target accessory to be detected, thereby effectively solving the problem of difficult installation of the traditional vibration sensor and avoiding the risk of fault identification of staff in a severe environment; in addition, the invention adopts a dislocation superposition method to enhance signals before feature extraction, improves signal to noise ratio, highlights weak fault signals contained in background noise, is beneficial to improving fault diagnosis accuracy based on acoustic signals, reduces influence of redundant features on diagnosis accuracy after feature extraction, reduces dimension of the time domain feature matrix by using a principal component analysis method, and filters redundant features to obtain an effective feature matrix. The centrifugal pump fault diagnosis model trained by the method is used for carrying out centrifugal pump fault diagnosis, and the type of the centrifugal pump fault can still be accurately identified under the interference of strong background noise.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description, serve to explain the principles of the application. It is apparent that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained from these drawings without inventive effort for a person of ordinary skill in the art. In the drawings:
FIG. 1 is a schematic diagram of a probabilistic neural network architecture shown in an exemplary embodiment of the present application;
FIG. 2 is a flow chart diagram illustrating a method of training a centrifugal pump failure diagnostic model according to an exemplary embodiment of the present application;
FIG. 3 is a flow chart illustrating a method of obtaining a first effective time domain matrix according to an exemplary embodiment of the present application;
FIG. 4 is a schematic flow diagram of a Harishawk algorithm optimized probabilistic neural network shown in an exemplary embodiment of the present application;
FIG. 5 is a flow chart diagram illustrating a method of diagnosing a failure of a centrifugal pump according to an exemplary embodiment of the present application;
FIG. 6 is a schematic diagram of a centrifugal pump failure diagnosis model training apparatus according to an exemplary embodiment of the present application;
fig. 7 is a schematic structural view of a centrifugal pump failure diagnosis device shown in an exemplary embodiment of the present application;
FIG. 8 is a detailed flow diagram of a method for training a fault diagnostic model of a centrifugal pump according to an exemplary embodiment of the present application;
FIG. 9 is a graph showing the number of iterations of the Harris eagle algorithm versus model verification error in accordance with an exemplary embodiment of the present application;
fig. 10 is a schematic diagram showing the diagnostic result of a centrifugal pump failure diagnosis model according to an exemplary embodiment of the present application.
Detailed Description
In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention. In addition, the drawings provided in the following embodiments merely schematically illustrate the basic idea of the present invention, and only the components related to the present invention are shown in the drawings, not according to the number, shape and size of the components in actual implementation, and the form, number and proportion of each component in actual implementation may be arbitrarily changed, and the layout of the components may be more complex.
It is noted that the terms "comprises" and "comprising," and any variations thereof, in the description and claims of the present invention and in the foregoing figures, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.
The following describes each technical term in the present application:
dislocation superposition method: the method can obtain the target signal under the complex working condition, can extract the required related components from the complex noise signal, and has the advantages of high calculation efficiency, high precision, improved experimental data processing efficiency and the like. The dislocation superposition method only has time domain calculation, and can improve the signal-to-noise ratio of the signal without damaging the required signal through the periodical superposition of the actual signal. The method avoids the mode mixing problem of a plurality of noise reduction methods, can keep specific frequency details and can also keep abnormal mutation information related to fault frequency. In the actual signal processing process, the superposition length is constant, so that a plurality of superposition end effectors can be caused, and in order to accurately extract the characteristics, a window function can be used for extracting the middle part of the superposition signal for subsequent processing. The effect of the dislocation superposition processing is related to the actual signal start position, and the optimum start position of the actual signal needs to be determined. Theoretically, the larger the superposition number K value of the dislocation superposition method is, the more accurate the waveform is, and the better the noise reduction effect is. However, as the K value increases, the more time is consumed and the noise reduction effect tends to be stable, so when the superimposed signal reaches the target requirement, the K value is determined. The algorithm only involves addition operation, so that the calculation is rapid, and the obtained signal model is more accurate. The specific formula is as follows:
where S (n) is the original signal,
is the DSM-processed signal, i.e. the superimposed signal, K is the number of overlaps, k=0, 1, 2.
The signal enhancement effect is evaluated by using the signal-to-noise ratio and the root mean square error, and the higher the signal-to-noise ratio or the smaller the root mean square error is, the better the signal enhancement effect is, and the expression is as follows:
where RMSE is root mean square error and SNR is signal to noise ratio.
Principal component analysis: the method is one of the most commonly used dimension reduction methods, and converts a group of variable data with possible correlation into a group of linear uncorrelated variables through positive-negative conversion, wherein the converted variables are called principal components, and the method comprises the following specific steps of:
representing the sample signal to be input q= [ Q ] with a vector matrix 1 Q 2 ···Q n ]The covariance matrix is:
where n is the number of training samples,
is the sample vector mean; q (Q) l Is the first column of the vector matrix Q;
calculating the eigenvalue and corresponding eigenvector of the vector matrix, and setting the eigenvalue lambda of the vector matrix i (i=1, 2, …, n), in order of magnitude, i.e. λ 1 >λ 2 >…>λ n Sample Q j The projection, its corresponding principal component amounts are:
wherein v is i The feature vector is the feature vector corresponding to the feature value;
constructing orthogonal space from the obtained eigenvectors, the principal component amount can be obtained by sample Q j Projection to orthogonal space, and in reconstruction, the contribution rate of each feature vector is proportional to the corresponding feature value, and the first n principal components in the orthogonal space are defined as y 1 ,y 2 ,…,y l The cumulative contribution rate is:
the preset value of θ is generally 95%, that is, when the contribution rate contained in the principal elements l (l < n) is accumulated to more than 95% of the information containing the original data, the principal elements can be used for representing the original information, so that the purpose of converting the high-dimensional data into the low-dimensional data is achieved.
Harris eagle algorithm: the bionic optimization algorithm for simulating the haris eagle group collaborative predation behavior has the advantages of outstanding global searching capability, high convergence speed and the like, and comprises the following three stages:
during the seeking phase, harris eagle predates by two strategies:
wherein X is rand Is the current position of a randomly selected individual of eagle group, X rabbit Is the current position of the prey, X m Is the average distance of the population, ub and l b are the upper and lower bounds of the global search range, r 1 、r 2 、r 3 、r 4 Is a random between (0, 1)A machine number;
in the blocking stage, the Harris eagle is changed from the whole domain into the local search:
wherein E is escape energy, t is iteration number at the time, E 0 Is a random number (-1, 1), T is the maximum number of iterations;
in the attack phase, harris eagle predates with 4 of 7 predatory strategies:
when 1 is more than or equal to |E| is more than or equal to 0.5 and r is more than or equal to 0.5 Harris hawk, a soft attack mode of wandering consumption is adopted:
X(t+1)＝ΔX(t)-EJX rabbit (t)-X(t)|
ΔX(t)＝X rabbit (t)-X(t)
wherein DeltaX (t) is the position difference, r 5 A random number of (0, 1), J is the jump strength of the hunting escape, and is a random number between (0, 2);
when |E| < 0.5 and r is larger than or equal to 0.5, the hunting escape failure is caught:
X(t+1)＝X rabbit (t)-E|ΔX(t)|
when 1 > |E| is not less than 0.5 and r is less than 0.5, the Harris eagle adopts a progressive rapid diving soft surrounding strategy:
Y＝X rabbit (t)-E|JX rabbit (t)-X(t)|
Z＝Y+S*LF(D)
wherein D is the dimension of the objective function, S is the D-dimension random vector, and LF is the Levy function;
when |E| < 0.5 and r < 0.5, the Harris eagle adopts progressive rapid dive hard enclosure:
Y＝X rabbit (t)-EJX rabbit (t)-X m (t)
Z＝Y+S*LF(D)
probabilistic neural network: the method belongs to a parallel algorithm of a radial basis function neural network, and has the advantages of good robustness, fast convergence, good compatibility and the like. The model can solve the problem of the nonlinear algorithm by using the linear algorithm, and the high-precision characteristic of the nonlinear algorithm is reserved, so that the neural network is more applied to fault diagnosis.
Fig. 1 is a schematic diagram of a probabilistic neural network according to an exemplary embodiment of the present application.
Referring to fig. 1, the probabilistic neural network may include an input layer, a pattern layer, a summation layer, and an output layer. The probability neural network input layer receives sample data from the outside of the structure, the number of neurons of the probability neural network input layer is the same as the length of input feature vectors, and the input layer carries out scalar product operation on the feature vectors and the weights of training samples to obtain a result; the probability neural network mode layer is a radial base layer, and the function of the layer is that the matching relation training calculation is carried out, namely, the matching relation between the input characteristic vector and each mode of the training set is used for calculating the Euclidean distance between a sample to be tested and the training sample; the nature of the probability neural network summation layer is an adder, and the layer sums the probabilities belonging to a certain type of modes through a kernel density estimation principle, namely, carries out weighted average processing on the outputs of similar neurons; the probability neural network output layer is composed of competing neurons, the essence of the layer is a comparator, and a Bayesian classification rule is adopted to make maximum posterior probability selection for fault types.
In order to solve the problem that the failure type of the centrifugal pump cannot be diagnosed efficiently and accurately in the prior art, embodiments of the present application respectively propose a centrifugal pump failure diagnosis model training method, a centrifugal pump failure diagnosis model training device, and a centrifugal pump failure diagnosis device, and these embodiments will be described in detail below.
Referring to fig. 2, fig. 2 is a flowchart illustrating a training method of a fault diagnosis model of a centrifugal pump according to an exemplary embodiment of the present application. As shown in fig. 2, in an exemplary embodiment, the centrifugal pump fault diagnosis model training method at least includes steps S210 to S230, which are described in detail as follows:
in step S210, centrifugal pump acoustic radiation signals in different states are acquired.
In a specific exemplary embodiment, the acquisition of the acoustic radiation signals of the centrifugal pump in different states requires suspending the microphone at a position 10 to 50 cm above the bearing of the driving end of the centrifugal pump, collecting 400 groups of acoustic radiation signals of the centrifugal pump in different states under the same working condition by the microphone, and dividing the data sets of the centrifugal pump in each fault type into a training set and a test set, wherein the different states comprise a normal state, misalignment of the shaft axis of the motor and the shaft axis of the centrifugal pump, loosening of the anchor bolts of the motor and the centrifugal pump, cavitation of the centrifugal pump, and 400 groups of sample working conditions and fault types are shown in the following table:
it should be noted that, in the conventional fault diagnosis method for the centrifugal pump, the vibration signal of the centrifugal pump is measured by a contact installation mode to perform diagnosis, and the vibration signal is often required to be installed on the surface of equipment to obtain the signal; and the acquisition of the acoustic radiation signal can realize non-contact measurement only by arranging the microphone near the target to be measured (particularly when the contact installation cannot be carried out in the high-temperature environment). Therefore, the non-contact signal acquisition mode based on the acoustic radiation signals can effectively solve the problem of difficult installation of the traditional vibration sensor, and avoid the risk of operation of staff in dangerous environments.
In step S220, the centrifugal pump acoustic radiation signal is enhanced, and the enhanced centrifugal pump acoustic radiation signal is processed to obtain a first effective time domain feature matrix.
It should be noted that, in the operation process of the centrifugal pump, the time domain characteristics of the acoustic radiation signals of the centrifugal pump change along with different fault states, and compared with the frequency domain characteristics and the time-frequency characteristics, the time domain characteristics have higher instantaneity, so that the time domain characteristics of the acoustic radiation signals are selected as input.
Illustratively, enhancing the centrifugal pump acoustic radiation signal, processing the enhanced centrifugal pump acoustic radiation signal to obtain a first effective time domain feature matrix includes: the dislocation superposition method is used for carrying out signal enhancement on the acoustic radiation signal, so that the signal-to-noise ratio of the acoustic radiation signal is improved, and fault signals under strong background noise are projected; calculating time domain characteristics of the sound radiation signals after signal enhancement, including average value, peak value, variance, root mean square value, crest factor, margin factor, pulse coefficient, shape coefficient, skewness and kurtosis, and then constructing a time domain characteristic matrix; and (3) performing dimension reduction on the time domain feature matrix by using a principal component analysis method, filtering redundant features, and normalizing the matrix to avoid the influence of dimension and signal amplitude, wherein the normalized feature matrix is used as a first effective time domain feature matrix. And processing the sample data of the centrifugal pump subjected to signal enhancement in sequence according to the steps, and dividing the processed data set into a training set and a testing set so as to train and test the probabilistic neural network later, wherein the training set and the data in the testing set have no cross overlapping part.
Before feature extraction, the invention adopts a dislocation superposition method to carry out signal enhancement on the acoustic signals, improves the signal to noise ratio, highlights weak fault signals contained in background noise, is beneficial to improving the fault diagnosis accuracy based on the acoustic radiation signals, reduces the influence of redundant features on the diagnosis accuracy after feature extraction, reduces the dimension of the time domain feature matrix by using a principal component analysis method, filters the redundant features, and obtains an effective feature matrix so as to facilitate the training and testing of a subsequent model.
In step S230, the first effective time domain feature matrix is used as a sample data set, and a probabilistic neural network is optimized through a harris eagle algorithm, so as to form a centrifugal pump fault diagnosis model.
The centrifugal pump fault diagnosis model is built, a data set sample is taken as input, the smoothing factors of the probabilistic neural network are selected in a self-adaptive mode by using a Harris eagle algorithm, as shown in fig. 9, when the 10 th iteration is carried out, the verification error is reduced to a lower value, and the network convergence requirement is met; then, referring to the fault diagnosis classification effect of the multi-classification confusion matrix inspection model, the test set of the sample is input into the model to inspect the accuracy of the model, as shown in fig. 10, the accuracy of fault identification reaches 100% in an exemplary embodiment, which indicates that the diagnosis effect of the model is better, and the method has a certain meaning for risk early warning of the fault of the centrifugal pump.
It should be noted that, to verify the effect of optimizing the probabilistic neural network using the harris eagle algorithm, in a specific embodiment, the probabilistic neural network optimized by the algorithm is compared with the inverse error neural network, the learning vector quantization neural network, and the conventional probabilistic neural network. The same training set is used for inputting the different neural networks for training, and the test set is used for verifying the trained model. The reverse error neural network parameters were set as follows: the maximum iteration number is 1000, the expected error of the network is 0.001, and the learning rate alpha is 0.01; the learning vector quantization neural network parameters are set as follows: the maximum iteration times are 500, the expected error of the network is 0.001, and the learning rate alpha is 0.1; the smoothing factors of the traditional probabilistic neural network are randomly set, the fault classification effect is shown in fig. 10, and the average accuracy of the test set is shown in the following table:
as can be seen from fig. 10 and the table above, by comparing the classification accuracy of the four models, the classification accuracy of the probability neural network model optimized by the harris eagle algorithm is better than that of the other three methods, and the fault diagnosis accuracy is better.
The method is characterized in that the probability neural network is optimized through the Harris eagle algorithm, the self-adaptive selection of the optimal smoothing factor is realized, the problem that the classification result is uncertain due to manual adjustment of the smoothing factor of the traditional probability neural network is solved, the convergence speed and the performance of the optimized network are good, and the reliable fault diagnosis of the centrifugal pump is realized.
As shown in fig. 3, in an exemplary embodiment, the method for obtaining the first effective time domain matrix includes at least steps S310 to S340, which are described in detail as follows:
in step S310, calculating a time domain feature of the enhanced centrifugal pump acoustic radiation signal;
in step S320, the time domain feature is constructed as a time domain feature matrix;
in step S330, the dimension of the time domain feature matrix is reduced by using a principal component analysis method;
in step S340, the time domain feature matrix after the dimension reduction is normalized, so as to obtain a first effective feature matrix.
As shown in fig. 4, in an exemplary embodiment, the harris eagle algorithm optimizing probabilistic neural network includes at least steps S410 to S470, which are described in detail as follows:
in step S410, a probabilistic neural network is established, a training set in the sample data set is input, a neural network model is trained, and the number of neurons in an input layer, a mode layer and a summation layer is initialized;
in step S420, initializing the parameters of the hawk algorithm, and setting the initial population number, the maximum iteration number, the upper and lower bounds of the global search range, and the error;
in step S430, calculating fitness values of each hawk individual in the population according to the upper and lower boundaries of the global search range, and sorting to determine that the hawk at the optimal position is the first generation hawk;
it should be noted that the above sorting method is to sort according to the principle that the smaller the fitness value is, the better the current position of the harris eagle is.
In step S440, the updating strategy classification is performed according to the escape energy factor and the value range of the random number, and iterative updating is performed on the harris eagle position according to different strategies;
in step S450, calculating the updated position of the harris eagle as the fitness after the iteration, comparing with the harris eagle at the optimal position of the previous generation, and retaining the harris eagle at the optimal position;
in step S460, when the fitness value reaches a set value or the iteration number of the algorithm reaches a preset value, training is terminated, otherwise, the harris eagle position is continuously updated;
in step S470, a final harris eagle position is obtained, and the harris eagle position is used as an optimal smoothing factor of the probabilistic neural network model.
As shown in fig. 5, in an exemplary embodiment, the method for diagnosing a fault of a centrifugal pump at least includes steps S510 to S530, which are described in detail as follows:
in step S510, collecting a centrifugal pump acoustic radiation signal;
in step S520, the centrifugal pump acoustic radiation signal is enhanced, the time domain feature of the enhanced centrifugal pump acoustic radiation signal is extracted, a time domain feature matrix is constructed, and the time domain feature matrix is subjected to dimension reduction clustering to obtain a second effective time domain feature matrix;
in step S530, the second effective time domain feature matrix is input to a centrifugal pump fault diagnosis model for performing fault diagnosis, where the centrifugal pump fault diagnosis model is obtained by training by the centrifugal pump fault diagnosis model training method shown in fig. 2.
As shown in fig. 6, in an exemplary embodiment, a centrifugal pump failure diagnosis model training apparatus 600 includes:
sample acquisition module 610: the method is used for acquiring acoustic radiation signals of the centrifugal pump under different states;
sample enhancement module 620: the method comprises the steps of enhancing the centrifugal pump acoustic radiation signal, and processing the enhanced centrifugal pump acoustic radiation signal to obtain a first effective time domain feature matrix;
model training module 630: and the first effective time domain feature matrix is used as a sample data set, and a probabilistic neural network is optimized through a Harris eagle algorithm so as to form a centrifugal pump fault diagnosis model.
It is not difficult to find that the centrifugal pump failure diagnosis model training apparatus embodiment is an apparatus embodiment corresponding to the centrifugal pump failure diagnosis model training method, and the centrifugal pump failure diagnosis model training apparatus embodiment may be implemented in cooperation with the centrifugal pump failure diagnosis model training method embodiment. The related technical details mentioned in the embodiment of the method for training the fault diagnosis model of the centrifugal pump are still valid in the embodiment of the device for training the fault diagnosis model of the centrifugal pump, and are not repeated here for the sake of reducing repetition. Accordingly, the related technical details mentioned in the embodiment of the centrifugal pump failure diagnosis model training apparatus may also be applied in the embodiment of the centrifugal pump failure diagnosis model training method.
As shown in fig. 7, in an exemplary embodiment, a centrifugal pump failure diagnosis apparatus 700 includes:
signal acquisition module 710: the device is used for collecting the acoustic radiation signals of the centrifugal pump;
signal enhancement module 720: enhancing the centrifugal pump acoustic radiation signal, extracting the time domain feature of the enhanced centrifugal pump acoustic radiation signal, constructing a time domain feature matrix, and performing dimension reduction clustering on the time domain feature matrix to obtain a second effective time domain feature matrix;
fault diagnosis module 730: the second effective time domain feature matrix is used for inputting the second effective time domain feature matrix into a centrifugal pump fault diagnosis model for fault diagnosis, wherein the centrifugal pump fault diagnosis model is obtained through training by the centrifugal pump fault diagnosis model training method shown in fig. 2.
It is not difficult to find that the centrifugal pump failure diagnosis apparatus embodiment is an apparatus embodiment corresponding to the centrifugal pump failure diagnosis method, and the centrifugal pump failure diagnosis apparatus embodiment may be implemented in cooperation with the centrifugal pump failure diagnosis method embodiment. The related technical details mentioned in the embodiments of the fault diagnosis method for the centrifugal pump are still valid in the embodiments of the fault diagnosis device for the centrifugal pump, and are not repeated here for the sake of reducing repetition. Accordingly, the related technical details mentioned in the embodiments of the centrifugal pump failure diagnosis apparatus can also be applied in the embodiments of the centrifugal pump failure diagnosis method.
As shown in fig. 8, in an exemplary embodiment, the specific flow of the centrifugal pump fault diagnosis model training method is as follows:
firstly, acquiring a centrifugal pump acoustic radiation signal, processing to obtain a feature matrix as sample data, and dividing the sample data into a training set and a testing set; initializing the Harris eagle algorithm parameters, calculating an fitness function, and calculating escape energy factors; when the stopping condition is not met, returning to the step of calculating the fitness function until the stopping condition is met; when the stopping condition is met, creating a probabilistic neural network, and training the probabilistic neural network by taking the escape energy factor as a smoothing factor of the optimized probabilistic neural network; when the iteration times reach a preset value or the error reaches an expected error, inputting the training set into a probability neural network to classify fault types; and finally, inputting the test set into a probabilistic neural network to test the centrifugal pump fault diagnosis model.
In summary, the invention discloses a centrifugal pump fault diagnosis model training method, a centrifugal pump fault diagnosis method and a centrifugal pump fault diagnosis device, which provides an improved probabilistic neural network model, can realize the self-adaptive selection of the optimal smoothing factor through the Harris hawk algorithm optimization, solves the problem that the classification result is uncertain due to manual adjustment of the smoothing factor of the traditional probabilistic neural network, and has better convergence speed and performance after optimization, thus realizing the reliable fault diagnosis of the centrifugal pump; in addition, the invention constructs a non-contact fault diagnosis method and system based on acoustic radiation signals, and only needs to arrange the microphone on the target accessory to be detected, thereby effectively solving the problem of difficult installation of the traditional vibration sensor and avoiding trouble of fault identification of staff in a severe environment; in addition, the invention adopts a dislocation superposition method to enhance signals before feature extraction, improves signal to noise ratio, highlights weak fault signals contained in background noise, is beneficial to improving fault diagnosis accuracy based on acoustic signals, reduces influence of redundant features on diagnosis accuracy after feature extraction, reduces dimension of the time domain feature matrix by using a principal component analysis method, and filters redundant features to obtain an effective feature matrix. The centrifugal pump fault diagnosis model trained through the method can accurately identify the fault type of the centrifugal pump under the interference of strong background noise, and has certain significance for risk early warning of the centrifugal pump fault.
The above embodiments are merely illustrative of the principles of the present invention and its effectiveness, and are not intended to limit the invention. Modifications and variations may be made to the above-described embodiments by those skilled in the art without departing from the spirit and scope of the invention. It is therefore intended that all equivalent modifications and changes made by those skilled in the art without departing from the spirit and technical spirit of the present invention shall be covered by the appended claims.
Claims (10)
1. A method for training a fault diagnosis model of a centrifugal pump, the method comprising:
acquiring sound radiation signals of the centrifugal pump under different states;
enhancing the centrifugal pump acoustic radiation signal, and processing the enhanced centrifugal pump acoustic radiation signal to obtain a first effective time domain feature matrix;
and taking the first effective time domain feature matrix as a sample data set, and optimizing a probability neural network through a Harris eagle algorithm to form a centrifugal pump fault diagnosis model.
2. The method for training a fault diagnosis model of a centrifugal pump according to claim 1, wherein the step of acquiring the acoustic radiation signals of the centrifugal pump in different states comprises:
disposing a microphone adjacent the centrifugal pump;
and setting different fault types of the centrifugal pump, and collecting centrifugal pump acoustic radiation signals in different fault states and centrifugal pump acoustic radiation signals in a normal state through the microphone.
3. A method of training a diagnostic model of a centrifugal pump according to claim 2, wherein said different fault types include: the motor shaft is not centered with the shaft axis of the centrifugal pump, the anchor bolts of the motor and the centrifugal pump are loosened, the centrifugal pump is cavitation corroded, the rolling bearing of the centrifugal pump is failed, the mechanical seal of the centrifugal pump is leaked, and the centrifugal pump is in coupling failure with the motor.
4. The method of claim 1, wherein said enhancing said centrifugal pump acoustic radiation signal comprises:
and carrying out signal enhancement on the centrifugal pump acoustic radiation signal by using a dislocation superposition method.
5. The method of claim 1, wherein processing the enhanced acoustic radiation signal of the centrifugal pump to obtain a first effective time domain feature matrix comprises:
calculating the time domain characteristics of the enhanced centrifugal pump acoustic radiation signals;
constructing the time domain feature as a time domain feature matrix;
performing dimension reduction on the time domain feature matrix by using a principal component analysis method;
and carrying out normalization processing on the time domain feature matrix after the dimension reduction to obtain a first effective feature matrix.
6. The method of claim 5, wherein the time domain features comprise: average, peak, variance, root mean square, crest factor, margin factor, pulse coefficient, shape coefficient, skewness, and kurtosis.
7. The method for training a fault diagnosis model of a centrifugal pump according to claim 1, wherein the optimizing the probabilistic neural network by the harris eagle algorithm using the first effective time domain feature matrix as a sample data set specifically comprises:
establishing a probabilistic neural network, inputting a training set in the sample data set, training a neural network model, and initializing the neuron numbers of an input layer, a mode layer and a summation layer;
initializing parameters of a Harris eagle algorithm, and setting initial population quantity, maximum iteration times, upper and lower boundaries of a global search range and errors;
calculating the fitness value of each Harisc eagle individual in the population according to the upper bound and the lower bound of the global search range, and sequencing to determine that the Harisc eagle at the optimal position is the first generation Harisc eagle;
performing updating strategy classification according to the escape energy factors and the value range of the random numbers, and performing iterative updating on the Harris eagle position according to different strategies;
calculating the updated position of the harris eagle as the fitness after the iteration, comparing the updated position with the harris eagle at the optimal position of the previous generation, and reserving the harris eagle at the optimal position;
when the fitness value reaches a set value or the iteration number of the algorithm reaches a preset value, training is stopped, otherwise, the Harris eagle position is continuously updated;
and obtaining the final Harris eagle position, and substituting the Harris eagle fitness value into a model as an optimal smoothing factor of the probabilistic neural network model.
8. A method of diagnosing a failure of a centrifugal pump, the method comprising:
collecting a centrifugal pump acoustic radiation signal;
enhancing the centrifugal pump acoustic radiation signal, extracting the time domain feature of the enhanced centrifugal pump acoustic radiation signal, constructing a time domain feature matrix, and performing dimension reduction clustering on the time domain feature matrix to obtain a second effective time domain feature matrix;
inputting the second effective time domain feature matrix into a centrifugal pump fault diagnosis model for fault diagnosis, wherein the centrifugal pump fault diagnosis model is trained by the centrifugal pump fault diagnosis model training method according to any one of claims 1 to 7.
9. A centrifugal pump failure diagnosis model training device, characterized by comprising:
sample acquisition module: the method is used for acquiring acoustic radiation signals of the centrifugal pump under different states;
sample enhancement module: the method comprises the steps of enhancing the centrifugal pump acoustic radiation signal, and processing the enhanced centrifugal pump acoustic radiation signal to obtain a first effective time domain feature matrix;
model training module: and the first effective time domain feature matrix is used as a sample data set, and a probabilistic neural network is optimized through a Harris eagle algorithm to form a centrifugal pump fault diagnosis model for centrifugal pump fault diagnosis.
10. A centrifugal pump failure diagnosis device, characterized by comprising:
the signal acquisition module: the device is used for collecting the acoustic radiation signals of the centrifugal pump;
a signal enhancement module: enhancing the centrifugal pump acoustic radiation signal, extracting the time domain feature of the enhanced centrifugal pump acoustic radiation signal, constructing a time domain feature matrix, and performing dimension reduction clustering on the time domain feature matrix to obtain a second effective time domain feature matrix;
and a fault diagnosis module: the second effective time domain feature matrix is used for inputting the second effective time domain feature matrix into a centrifugal pump fault diagnosis model for fault diagnosis, wherein the centrifugal pump fault diagnosis model is trained by the centrifugal pump fault diagnosis model training method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310417097.5A CN116383661A (en) | 2023-04-13 | 2023-04-13 | Centrifugal pump fault diagnosis model training method, fault diagnosis method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310417097.5A CN116383661A (en) | 2023-04-13 | 2023-04-13 | Centrifugal pump fault diagnosis model training method, fault diagnosis method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116383661A true CN116383661A (en) | 2023-07-04 |
Family
ID=86967359
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310417097.5A Pending CN116383661A (en) | 2023-04-13 | 2023-04-13 | Centrifugal pump fault diagnosis model training method, fault diagnosis method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116383661A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118013289A (en) * | 2024-04-09 | 2024-05-10 | 北京理工大学 | Variable working condition small sample fault diagnosis method, device, medium and product based on information fusion element transfer learning |
-
2023
- 2023-04-13 CN CN202310417097.5A patent/CN116383661A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118013289A (en) * | 2024-04-09 | 2024-05-10 | 北京理工大学 | Variable working condition small sample fault diagnosis method, device, medium and product based on information fusion element transfer learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109164343B (en) | Transformer fault diagnosis method based on characteristic information quantization and weighted KNN | |
| CN110792563B (en) | Wind turbine generator blade fault audio monitoring method based on convolution generation countermeasure network | |
| CN110617966A (en) | Bearing fault diagnosis method based on semi-supervised generation countermeasure network | |
| Chen et al. | A meta-learning method for electric machine bearing fault diagnosis under varying working conditions with limited data | |
| CN111832353B (en) | Steam turbine rotor fault diagnosis method based on EMD and BA optimization SVM | |
| Li et al. | Bearing fault feature selection method based on weighted multidimensional feature fusion | |
| CN110132554B (en) | Rotary machine fault diagnosis method based on deep Laplace self-coding | |
| Guo et al. | A new fault diagnosis classifier for rolling bearing united multi-scale permutation entropy optimize VMD and cuckoo search SVM | |
| CN112729834B (en) | Bearing fault diagnosis method, device and system | |
| Chen et al. | Intelligent model-based integrity assessment of nonstationary mechanical system | |
| CN116226646A (en) | Method, system, equipment and medium for predicting health state and residual life of bearing | |
| CN111043050A (en) | Fault diagnosis method and system for centrifugal pump | |
| Afrasiabi et al. | Two-stage deep learning-based wind turbine condition monitoring using SCADA data | |
| Zhao et al. | A novel deep fuzzy clustering neural network model and its application in rolling bearing fault recognition | |
| CN114118219A (en) | Data-driven real-time abnormal detection method for health state of long-term power-on equipment | |
| CN115901249A (en) | Rolling bearing performance degradation evaluation method combining feature optimization and multi-strategy optimization SVDD | |
| CN115791174A (en) | Rolling bearing abnormity diagnosis method and system, electronic equipment and storage medium | |
| CN115345255A (en) | Fault diagnosis method, control device, terminal and storage medium | |
| Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
| CN115081321A (en) | Corrosion fatigue life prediction method, system and equipment for marine welding structure | |
| CN113126489A (en) | CNN-GRU-BINN-based intelligent BIT design method for heavy-duty gas turbine control system | |
| CN112287752A (en) | Method for extracting early fault characteristics of rotating shaft of hydroelectric generator | |
| CN116792264A (en) | GRU-based wind driven generator multi-fault diagnosis method | |
| CN116383661A (en) | Centrifugal pump fault diagnosis model training method, fault diagnosis method and device | |
| CN116644348A (en) | Cross-mechanical part fault diagnosis method and device based on transfer type countermeasure migration |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |